|Publication number||US5498990 A|
|Application number||US 08/417,511|
|Publication date||Mar 12, 1996|
|Filing date||Apr 5, 1995|
|Priority date||Nov 5, 1991|
|Also published as||US5737587|
|Publication number||08417511, 417511, US 5498990 A, US 5498990A, US-A-5498990, US5498990 A, US5498990A|
|Inventors||Wingyu Leung, Winston Lee, Fu-Chieh Hsu|
|Original Assignee||Monolithic System Technology, Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (76), Non-Patent Citations (31), Referenced by (151), Classifications (47), Legal Events (5)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is a continuation of U.S. patent application Ser. No. 08/156,186, filed Nov. 22, 1993, now abandoned, which in turn is a continuation-in-part of U.S. patent application Ser. No. 07/927,564 filed Aug. 10, 1992, entitled "Fault-tolerant, High Speed Bus System and Bus Interface for Wafer Scale Integration", now abandoned, which, in turn is a continuation-in-part of U.S. patent application Ser. No. 07/865,410 filed Apr. 8, 1992 entitled "Circuit Module Redundancy Architecture", now abandoned, which, in turn is a continuation-in-part of U.S. patent application Ser. No. 07/787,984 filed Nov. 5, 1991 entitled "Wafer-scale Integration Architecture, Circuit, Testing and Configuration", now abandoned.
1. Field of the Invention
This invention relates to a low latency, high bandwidth bus system for use in a high-performance, large capacity memory system, logic system or combined logic and memory system, and particularly to such a bus system suitable for conventional VLSI devices or wafer-scale integration devices with a circuit module-oriented architecture.
2. Description of the Prior Art
Many traditional bus systems use reduced-swing signal transmission (i.e., a signal swing smaller than the supply voltage), including TTL standard, to enable high speed operations. Problems usually arise in data transmission when either of the following conditions exists: (i) the rise or fall time of the transmitted signal is a significant fraction of the clock period; (ii) there are reflections on the signal bus which interfere with the rising or falling transitions of the signal. The data transfer rate is limited in part by whether signal integrity is compromised as a result of the above conditions. The other hindrance is the maximum cycle rate of the memory and controller. To increase data bandwidth, the above conditions should be avoided.
High frequency data transmission through a bus requires a high rate of electrical charge (Q) transfer on and off the bus to achieve adequate rise and fall times. To avoid condition (i) above, large transistors in the memory and controller are needed to source and sink large amounts of current required to switch the signal levels. This requirement undesirably causes increased silicon area, bus capacitance, power consumption and power supply noise.
Assuming transmission rate is a constant, there are several known ways to improve the situation. One is to minimize the bus capacitance, but this severely constrains the bus length as well as the number of devices that can be attached to the bus. The second method is to reduce the voltage swing needed to distinguish between logical 1 and 0. This reduces the amount of charge transfer; thereby, power consumption, noise and silicon area can be reduced. To minimize condition (ii), impedance matching using proper line termination can be used to reduce reflection. The reference voltage, or trip point, of a signal bus is the voltage level which delineates a logical 1 from a logical 0. Most digital CMOS circuits use supply voltage Vdd/2 as the trip point, and this permits the circuits to operate with symmetric noise margins with respect to the power and ground supply voltages. This also results in symmetrical inverter output rise and fall times because the PMOS and NMOS transistors in the circuit have equal strengths due to the fact that Vdd/2 is the trip point. With the exception of center-tap termination (CTT), prior art small swing (<1.5 v peak-to-peak) I/O schemes generally have symmetric voltage swings about a reference voltage that has no intended relation to the "natural" implicit reference voltage (Vdd/2) about which typical digital CMOS circuit operates. Examples of these schemes are the Gunning-Transceiver Logic (GTL) (R. Foss et al, IEEE Spectrum Oct. 1992, p. 54-57, "Fast Interfaces for DRAMs"), Rambus DRAM (M. Farmwald et al, IEEE Spectrum Oct. 1992, p. 50-51 "A Fast Path to One Memory"; and R. Comeford et al, "Memory Catches Up", p. 34-35, same issue). These require an explicit voltage reference source, either internal or external to the chip. In order to attain the desired attributes of symmetric noise margins and rise and fall time, these schemes must translate their small swing I/O reference point to the "natural" reference. This translation incurs the penalty of additional circuit delay and complexity. Furthermore, each bus line is terminated to the reference supply through a resistor with a value matching bus line impedance.
Center-tap termination, (R. Foss et al, IEEE Spectrum Oct. 1992, p. 54-57, "Fast interfaces for DRAMs") differs in that this scheme can accommodate Vdd/2 as the I/O reference without the need for an external reference supply voltage. Bus line resistive termination to Vdd/2 can be attained through a resistive voltage divider coupled between Vdd and ground. Small swing signal is attained through a feedback mechanism which reduces the drive of the output driver once the output voltage crosses the termination reference voltage (Vdd/2).
There are several drawbacks to this scheme. First, the logic required to perform the feedback mechanism is directly in the data path. This means added data-out delay or latency. Also the transistor count and static DC power required to implement this feedback control logic, which includes a threshold detector, may become a burden. Finally, this scheme requires two termination resistors per bus line. If a relatively wide bus were to be employed, this results in high component count and increase in required board space.
In accordance with the present invention, a compact, high speed reduced CMOS swing I/O scheme uses the natural CMOS inverter trip point as its reference voltage, without an external reference voltage supply. Furthermore, the standby current present in prior art reduced-swing schemes can be eliminated during the standby mode when no data is transmitting in the present invention.
Prior art bus architectures, whether in ICs or in system boards, make no distinction, relative to signal swing amplitude, with respect to direction of signal travel across the bus. The signal swing amplitude transmitted from one end of the bus is identical to transmissions sent from the other direction. In accordance with the present invention, an asymmetrical signal-swing internal bus optimizes the speed, area and power of the bus system with few masters and many slave devices.
For high-speed burst mode operation, the prior art method to sequence the column select lines is to load the starting address into a counter controlled by the memory clock and connect the counter output to a decoder. The output of this decoder is buffered to drive the column select lines. At every rising clock edge the counter increments and its output decoded to generate the next active column select line. The column select lines are asserted in consecutive order for the duration of one clock cycle. The drawback to this approach is that the clock to valid column-select delay time is the sum of the clock to out time of the counter, the propagation delays of the decoder and the buffer. This delay directly affects the burst frequency and therefore the access bandwidth. In addition, since the decoder typically is implemented with combinational logic, the propagation delay through the decoder is not uniform for each input transition. As a result, simultaneous assertion of one or more column selects may occur for the duration of the decoder delay mismatches causing memory read or write failures. The present invention allows decoding of the address input first, followed by a barrel shifter circuit before driving the column select lines to eliminate the above mentioned shortcomings.
Prior art memory designs attempt to increase bandwidth by implementing synchronous burst transfers. Latency for the first accessed data is increased due to the insertion of a pipeline to increase the data rate. In these designs incorporating synchronous burst transfer mode, access latency is quantized and measured in clock cycles as opposed to traditional non-synchronous memory where access latency is measured in time units. (E.g., see data sheet from Samsung Electronics, "256K×4 units and bit Synchronous SRAM", and Motorola Semiconductor "32K×9 bit BurstRAM-Synchronous SRAM) The access latency (the time from the clock transition which latches the read command to the time when data is valid to the requestor) can only be resolved after one clock period regardless of the frequency of the clock. In accordance with the present invention, this shortcoming is reduced by using a resynchronization FIFO/buffer triggered by both the source clock and the destination clock signals.
Traditional memory devices are usually arranged in an x-y array with each of the dimensions being an integer power of 2. This arrangement does not, in general, fit with typical computer display resolutions (such as 640×480, 1024×768, 1280×1024, etc.) and, as a result wastes large number of memory bits that cannot be displayed. Many approaches have been proposed in the prior art to solve this problem, all of them in effect some kind of remapping of the display to fit into the memory device. This approach (for example, Y. Kudo et al, "Frame Buffer Memory for Display", U.S. Pat. No. 4,980,765), however, involves complicated calculation of pixel data movement during each display screen refresh and screen pixel update, and significantly reduces the performance of the display system. In accordance with the present invention, with circuit-module oriented architecture, a memory device (such as DRAM) is partitioned into many small blocks each with its own independent addressing means and dynamic address mapping capability, and thereby allows remapping the memory to fit the pixel arrangement of the display with minimum memory bits used.
Prior art multiple-write operations such as the "block-write" operation in dual-port DRAM allow simultaneous data written into multiple column locations (the sense amplifiers) from a pre-loaded color register. (e.g., see the data sheet for "1M bit dual-port DRAM", Mitsubishi Electric.) This arrangement allows multiple horizontal pixels in a display memory to be written with the same pixel data only. However, it cannot be used when varying data need to be written, or when multiple vertical pixels need to written with the same data such as during a pattern fill operation of the display memory. One disclosure uses multiple memory devices operated in an interleaved mode to improve this situation (e.g., G. Moffat, "Method and Apparatus for Arranging Access of VRAM to Provide Accelerated Writing of Vertical Lines to an Output Display", U.S. Pat No. 5,142,276). The present invention allows a broadcast-write operation that allows burst data stream to be written into selective, multiple memory circuit modules (that can be mapped into different vertical lines in a display memory) simultaneously and thereby dramatically increase write data bandwidth.
Existing memory, when burst mode is implemented, can only burst data out from the column circuitry (so called "page" mode, and is typically mapped into the horizontal direction in a display memory; see the above referenced Mitsubishi data sheet for dual-port DRAM). Any vertical pixel movement requires the memory to exit page mode, load another row of data and then access the next pixel data in the then current "page". This access is much slower than the horizontal access. One approach to solve the problem is to implement a special vector addressing scheme-horizontal vector and vertical vector, of the memory page or column circuitry (R. E. Harlin et al, "Dynamic Video RAM Incorporating On Chip Line Modification", U.S. Pat. No. 5,148,523). This approach, however, requires individual (or a few) bit-planes to be updated sequentially, i.e. relatively slow, and the dual-address decoding circuitry is complicated and area consuming. In accordance with the present invention, with its multiple block architecture, both "horizontal" burst (from the column circuitry of each block) and a inter-block burst mode (from consecutive, participating blocks in sequence) read/write operation are provided that can be mapped into the "vertical" pixel direction, therefore allowing high speed access of pixel data in both display directions.
In accordance with the present invention, a reduced CMOS-swing bus circuit includes clamping circuitry for each bus line. The clamping circuitry includes a feedback circuit sensing a signal on the bus line for generating a feedback signal, and a complementary source follower responsive to the feedback signal for clamping the signal on the bus line to provide the reduced CMOS-swing. The feedback circuit is a CMOS inverter and can be disabled to fix the signal on the bus line to a reference level when the bus line is inactive in a preferred embodiment.
A bus circuit for generating directional asymmetrical signal swing on a bus includes a first bus driver circuitry for generating a first signal swing for data transfer over said bus along a first bus direction, and a second bus driver circuitry for generating a second signal swing for data transfer over the bus along a second bus direction. The first signal swing is a full CMOS-swing and the second signal swing is a reduced swing substantially smaller than the first signal swing.
An address generation circuit includes an address decoder receiving a starting address of N bits for generating an address select signal on its 2N output lines and a barrel shifter clocking in said address select signal for shifting the clocked address select signal in a consecutive order.
A resynchronization circuit for resynchronizing a data signal with a source clock signal to a destination clock signal includes an input gate circuit for gating input data with a source clock signal, a buffer for receiving the gated data from the input gate circuit, and a output gate circuit for gating the data stored in the buffer with a destination clock signal. The buffer is a FIFO memory, and includes input and output select circuit in a preferred embodiment.
In accordance with the present invention, in a circuit system with a plurality of circuit modules and a bus connecting thereto, each circuit module has a unique preset address code, and at least one of circuit modules includes at least one set of software-programmable base address registers for mapping the circuit modules into an address space. The base address register has more address space than the preset address codes. The circuit modules are mapped into non-contiguous address space and the content of the base address register can be modified dynamically in some embodiments.
In accordance with the present invention, a scheme for simultaneously writing a data signal into selected subset blocks of a plurality of circuit blocks includes the steps of providing each circuit block with a broadcast-write mode register, issuing a broadcast-write-mode command on the bus for setting the broadcast-write status register in the selected subset of the plurality of circuit blocks, and broadcasting a write command so that circuit blocks with the broadcast-write-mode register set respond to the write command.
A method for inter-block burst operation across multiple circuit block includes providing a mode-bit in each circuit block, setting the mode-bit in circuit blocks participating the burst operation and accessing the participating circuit blocks at each clock cycle in a repeated sequence. The sequence to access the participating circuit blocks corresponds to a preset base address sequence.
Also in accordance with the present invention, a circuit subsystem includes a plurality of circuit blocks and a bus connecting thereto. At least one of the circuit blocks includes a redundant circuit, a test register for enabling testing of the redundant circuit, and a set of repair registers or fuses for setting the redundant circuit into operation. The test register and repair registers are programmable through bus commands.
Also in accordance with the present invention, an I/O bus in a memory subsystem includes multiplexed address and data I/O lines for carrying an address signal, and data signal, command input lines for carrying command signal, dynamic data mask input line for masking a part of the data signal on the bus from a data write operation, and a clock input line for carrying a clock signal. The address signal includes a base address field for addressing a memory module and an access address field for addressing a location or locations in the addressed memory module.
The above and other objects, features and advantages of the present invention will be clear from the following detailed description in conjunction with the accompanying drawings.
FIG. 1 shows a memory system in accordance with the present invention which is connected to various memory devices through a Reduced-CMOS-Swing Bus (RCSB).
FIG. 2(a) shows a reduced-CMOS-swing line driver with active source-side termination.
FIG. 2(b) shows a push-pull gate configuration which can replace the feedback inverter of FIG. 2(a).
FIG. 3 shows a memory device with its Internal Memory Bus (IMB) connecting the chip I/O area and memory modules.
FIG. 4 shows a memory module attached to the IMB through a Memory Bus Interface (MBI).
FIGS. 5a and 5b show prior art address mapping for display with (a) 1280×1024, (b) 1024×768 pixels.
FIGS. 6(a), 6(b) show an example of address mapping for displays with (a) 1280×1024, (b) 1024×768 pixels using block ID address mapping.
FIG. 7 shows a broadcast-write operation.
FIG. 8 shows a reduced-CMOS-swing sensing circuit with active destination-side termination.
FIG. 9 shows a prior art address sequencer circuit.
FIG. 10(a) shows an address sequencer circuit in accordance with the present invention.
FIG. 10(b) shows a barrel shifter consisting of 2N registers arranged in a ring.
FIG. 11 shows an inter-block burst read/write operation.
FIG. 12(a) shows a prior art resynchronization circuit.
FIG. 12(b) shows a simplified timing diagram of the resynchronization circuit in FIG. 12(a).
FIG. 13 shows a resynchronization circuit.
FIG. 14a and 14b show the simplified timing diagram of the circuit described in FIG. 13 with latency count (a) 4, first data out from rising edge; and (b) 3, first data out from falling edge.
FIG. 15(a) shows a memory block with test register, and redundant sub-arrays (row and column).
FIG. 15(b) shows a circuit diagram of the test register of FIG. 15(a).
A high-performance, high-density DRAM device using a circuit module-oriented architecture is described in this disclosure as an example that implements many of the novel features of this invention.
Bus Format and Protocol
FIG. 1 shows an embodiment of a memory system, in which memory devices (DRAMs) 101, 102 . . . are coupled to one (or more) bus masters (memory controllers) 111 through a reduced-CMOS-swing bus (RCSB) 150 which includes 28 signal lines: four dedicated command lines, sixteen multiplexed data and address lines, four data write mask lines for writing partial data to the memory (for example as shown in Table 1, DQM  masks data in D [3:0], DQM  masks data in D [7:4], . . . during write operation), one clock line, one chip select line, one output enable line and one clock enable line. Table 1 lists the 28 I/O pin definitions which include signal symbol, name, input/output direction and functional definition of a 1M Byte (8M bit) DRAM memory in one embodiment of the present invention. A typical command requires 20 bits of information: Four bits encode the operation to be performed, and depending on the contents of the four command bits, the remaining sixteen bits can be interpreted as a combination of the following: module address or identification code (ID), row address, column address, command-code extension or configuration data.
TABLE 1______________________________________Pin definitionSignal Name I/O Function______________________________________D[15:0] Data/address bus I/O Burst transfer dataC[3:0] Command bus I Command inputDQM[3:0] Dynamic data mask I Dynamic nibble maskCke Clock enable I Synchronous clock enableClk Clock I Burst transfer timingOE# Output enable I Output enable for burst readCS# Chip Select I Chip Enable______________________________________
Each command is issued referenced to a particular transition of the clock, in this case, a low to high transition. Data is grouped as half-words of 16 bits each. Single, consecutive half-words are transmitted at each clock transition, to or from the memory device during a burst write or burst read operation, respectively (this is called dual-edge transfer). Essentially, this allows a 32-bit word to be transmitted every clock cycle using the 16-bit data bus. Conventional data transfers are single edge transfers where a unit of data is transmitted at each rising or falling clock edge, but not at both.
The communication protocol between bus master (controller) and slave (memory) devices minimizes overall data access latency, bus width and operational clock frequency while maximizing data transfer bandwidth. Latency is reduced by simultaneously sending command code and all necessary information, including the full access address over the bus so that the desired command can be executed immediately. A scheme such as one described in Int. Pat No. PCT/US91/02590 [Farmwald et al] has a protocol which requires command and address to be multiplexed and sent at different clock edges; furthermore, because of the use of a narrow bus (8 bits), the address itself has to be segmented and multiplexed at different clock edges. With such a scheme, several clocks must transpire before the entire address is presented to the memory. This translates to latency in addition to the memory's inherent latency. The present system in contrast allows the overall latency to be very close to the inherent latency of the memory, because only one clock transition is needed for the entire address to be presented to the memory device.
Most memory access requests are directed at sequentially addressed memory locations. By taking advantage of this, multiple latency penalties incurred for multiple memory requests are reduced to one latency penalty by requiring only one request for sequential accesses. This is achieved by using a burst operation, which allows sequentially addressed data to be accessed given only the beginning address of a desired data stream. Burst operation eliminates the address generation bottleneck in the controller as well as in the address decoding delay in the memory, particularly in an embodiment with a novel address sequencing circuitry as described below. Burst mode operation also allows address and data to be multiplexed onto a common bus without compromising on data bandwidth. The ability to multiplex address and data significantly reduces the number of lines in the RCSB. This reduces memory pin count, power dissipation, PC board cost and complexity.
A burst read command is generated by establishing, with sufficient setup and hold time with respect to the rising edge of the clock, the burst read command-code on the four bit command bus. The module ID (the circuit block base address or identification code) and the beginning burst address is simultaneously established on the 16 bit multiplexed address/data bus. After some latency, data will stream from the memory module onto the multiplexed address/data bus, with each half-word transmitted at successive transitions of the clock (dual-edge transfer).
Achieving high bandwidth (>500 Mbytes/sec) while maintaining a relatively low operating frequency (>125 MHz) is accomplished by dual-edge transfer burst mode and a 16 bit data path. At this operating frequency, memory design and fabrication can be done with sufficient margin using existing processes. Running a bus at this clock rate on a PC (personal computer) board environment is possible by careful conventional PC board design, and attention to high frequency transmission issues.
Reduced Swing Bus
A compact, high speed Reduced CMOS Swing I/O scheme (RCSB) uses the natural CMOS inverter trip point as its reference voltage without any explicit reference voltage supply. Furthermore, the static DC current present in prior art reduced-swing schemes can be eliminated during the inactive mode in accordance with the present invention.
This is accomplished by the bus-line driver circuit shown in FIG. 2(a). The bus-line driver circuit 200 includes a line driver 201 which drives the bus line 205, and a feedback clamping circuit 202 which further includes a feedback circuit 203 (an inverter I2 in the illustrated embodiment) and a complementary source follower circuit 204. The bus line 205 is driven by a typical CMOS inverter 201 (I1); however, the line voltage is sensed by inverter 203 (I2) which drives the gate of feedback source follower transistors P2 and N2 with their sources coupled to the bus line 205. When the input of the inverter 201 (I1) is driven low, the line voltage (OUT) will go above the trip point of the inverter 203. As the output of the inverter 201 goes low, transistor N2's conductivity decreases while transistor P2's conductivity increases, thus preventing OUT from going to supply voltage. Conversely, if the input of the inverter 201 is driven high, transistor N2's conductivity increases and transistor P2's conductivity decreases to prevent OUT from reaching ground.
The peak-to-peak voltage swing as well as the common mode voltage generated by the bus line driver can be adjusted by varying the sizes of transistors P2, N2, P1, N1. The receiver for this signal transmission scheme is a conventional CMOS inverter. By using appropriate transistor sizes, the line driver output voltage swing can be centered to that of the receiver trip point.
The feedback inverter 203 (I2) in FIG. 2(a) may also be implemented using NAND gate, NOR gate, push-pull gate or other logic combinations with a control input to disable the feedback path (thereby forcing the inputs of the source follower transistors at either supply level jointly or separately to each supply level) when the bus is inactive to further reduce or eliminate static DC power dissipation. An example of push-pull gate configuration is shown in FIG. 2(b). Also, the feedback clamping circuit 202 can be implemented either in the driver side or the receiver side or both or on a separate chip.
Typical MOS output drivers are not suitable for high frequency transmission because of the excessive reflections caused by gross mismatches between the output impedance of the drivers and the line, or voltage bounces due to large switching transients. By adjusting the device size of the feedback device, the bus-line driver circuit 200 in FIG. 2 can be made suitable for driving typical PC board traces because this driver circuit has a predictable output impedance which is close to that of PC board traces. The inclusion of the feedback devices also serves to stabilize the output impedance of the driver circuit, as the drivers remain in the transistor linear region throughout the signal transition.
With this RCSB scheme, there is no need for internal or external voltage reference sources. Level translating circuits are not required since the reduced swing signal is referenced already to the inverter trip point; hence, translator circuitry and its associated delay is avoided. Furthermore, the inclusion of active bus line termination in the bus line driver circuit eliminates the need for external termination resistors.
The DRAM used in the present exemplary embodiment has a capacity of 9.4M bits. The memory is divided into 18 modules of 512K bits each. Sixteen of the memory modules are for regular use and two are spares. FIG. 3 shows a block diagram of the memory modules 301 to 318 connected together by an internal memory bus (IMB) 320. The internal memory bus 320 connects to outside circuitry through the chip interface 321. The chip IO interface 321 connects the memory modules 301 to 318 to the external memory bus 350 described above.
The memory modules 301 to 318 are identically constructed, each with a programmable base address (ID) so that one can replace another. The identification register used to store the module ID acquires its contents from either of two sources--pre-programmed (such as fuse programmed) ID, or controller issued ID. At power-up, each module is assigned a default communication ID previously specified through the use of fuses and/or other pre-programmed code. The memory module ID subsequently can be changed by writing-into the ID register through the use of an ID write command.
FIG. 4 shows the organization of a memory module 400 which contains 512K bit of memory divided into two banks 401, 402 each of 256 rows and 1024 columns. Each bank 401, 402 has its own independent row, column and sensing circuitry 405, 406. The latch-type sense amplifiers 405, 406 in each bank form a data buffer of 128 bytes. Read access to the memory is carried out in two steps. First, a row of data (1024 bits) addressed, using the RAS command, is transferred from the memory to corresponding sense amplifiers 405, 406. Next, one or multiple words (32 bits) of data addressed using the burst CAS command is transferred from the sense amplifiers 405, 406 to the internal bus. Thereafter, at each clock cycle the CAS address is incremented and the word associated with this address is transferred from the sense amplifiers to the internal bus. Using dual-edge transfer, a 16-bit bus can transfer a 32-bit word every clock cycle. Sixteen bits can be transferred to or from one of the two banks 401, 402 at the rising edge, and another 16 bits can be transferred to or from the other bank at the falling edge of the clock. Therefore, the column circuitry in each bank operates at the frequency of the memory clock as the bus.
Dynamic Address Mapping
The dynamic address mapping (ID mapping) capability of the memory modules and banks (collectively called blocks) in the present memory system with many (32 banks per 1 MByte memory in the present embodiment) blocks, especially when the mappable memory space is larger than the physical memory size, is particularly beneficial in applications such as virtual memory mapping or mapping display memory into direct display pixel organization. The latter provides the benefits of reduced memory bit requirement as well as much faster access and performance since no memory scrambling or re-formatting is required. FIG. 5 shows prior art display pixel mapping into display memory for (a) 1280×1024 pixel resolution and (b) 1024×768 pixel resolution. The former case requires memory size of 2048×1024 and the latter case requires 1024×1024 (assuming each pixel requires 1 byte of memory), both much bigger than the screen display size. In the present invention, as shown in FIGS. 6(a) and (b) as examples, the individual blocks can be mapped into non-contiguous address space but covering the entire display area with a much smaller number of bits required: 1280×1024 for FIG. 6(a), and 1024×768 for FIG. 6(b).
For example, for a 6 bit ID register (addresses up to 64 blocks from 0 to 63) configuration, the mapping in FIG. 6(a) can be set up as follows:
horizontally: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, then skipping 10 to 15
16, 17, 18, 19, 20, 21, 22, 23, 24, 25, then skipping 26 to 31
32, 33, 34, 35, 36, 37, 38, 39, 40, 41, then skipping 42 to 47
48, 49, 50, 51, 52, 53, 54, 55, 56, 57 then skipping 58 to 63.
Another example for the mapping in FIG. 6(b) only uses a 5 bit ID register:
horizontally: 0 to 7, 8 to 15, 16 to 23 then skipping 24 to 31.
Internal Bus Architecture
The memory module uses the internal memory bus (IMB) to communicate between the module bus interface (MBI) and the chip I/O. Sixteen multiplex address/data lines, four data write mask lines, four command lines, two clock lines (one from the chip I/O and one from the selected memory module in a source-synchronous fashion) and one output enable line comprise the IMB. The IMB employs a unique signaling architecture designated here Directionally Asymmetric Signal Swing (DASS), which minimizes bus line driver size and noise associated with driving large capacitive loads.
Existing bus architectures, whether in ICs or in system boards, make no distinction, relative to signal swing amplitude, with respect to direction of signal travel across the bus. The signal swing amplitude transmitted from one end of the bus is identical to that sent from the other direction. In the cases where there are many slave devices and one or few masters communicating through a large swing bus, the bus line drivers of both the slaves and the master must be large enough to drive the bus line at the operating frequency. (In this implementation, the slave devices are the memory modules, and the master is the chip I/O.) Large drivers in the memory modules consume more silicon area as well as increase bus line capacitance; furthermore, as the number of memory modules attached to the bus increases, bus capacitance increases. Eventually, increases in driver size does not result in increase of driving speed; consequently, the number of modules on the bus is severely limited. For example, in order to operate at 125 MHz only 6 to 8 modules can be attached to the 5 volt swing bus using a 1μ CMOS process. Large drivers needed to switch heavily loaded lines also significantly increase noise in the chip. Small swing buses, on the other hand, can be used to reduce power supply noise and the size of the bus line drivers, but small swing signal translators or amplifiers for each signal of the bus must be present in every memory module. In an implementation where many memory modules are placed in one die, this results in significant consumption of silicon area and power dissipation.
The DASS architecture in accordance with the present invention capitalizes on the benefits of using both large and small signal swings on the same bus. This architecture takes advantage of the asymmetry of communication in a master-slave communication environment where communication from master to slave is one-to-many (broadcast), but communication from slave to master is one-to-one. Large signal swing is used in the point-to-multiple destinations mode, when master is driving the bus, while small swing is used in the point-to-point mode when one of many memory devices drives the bus.
Executing a unique broadcast-write mode in the present implementation is a good example of a broadcast operation. In the broadcast-write operation, multiple memory banks can be written simultaneously with the same data. Data can be written simultaneously to any particular group of banks or to all banks. This is accomplished by first using a special command called "Broadcast-Write Select" which selects or deselects one or more memory module for participating in the subsequence write operation by setting (select) or resetting (deselect) a group of or all banks a broadcast-write status registers associated with banks in the memory chip. The broadcast-write select command takes one clock cycle. Once a bank is selected for broadcast write, it stays selected and participates in all subsequent write operation until it is deselected by the broadcast write select command. Next, a normal write command is broadcasted with all its write data following. All banks with their broadcast-write status register set will respond to the write command.
For example, in one embodiment, a broadcast-write select command can be used to select up to 4 memory banks. Eight broadcast-write commands back-to-back can be used to select 32 memory banks in the device. Then a normal memory write command can be executed to write a burst of data to all the selected banks simultaneously. In this example, eight cycles are used for the eight broadcast write commands and depends on the burst length of the write data; 2 to 33 cycles are used for writing the data and an additional cycle is required to terminate the burst write. FIG. 7 shows an example of a broadcast-write operation in the memory device.
Memory modules 701, 718 are selected for broadcast-write. Thus the broadcast-write status registers in the modules 701, 718 are set by the command. Memory module 702 is addressed by the address associated with the write command and data is therefore simultaneously written into the selected modules 701, 718 and the addressed module 702. The broadcast-write operation significantly increases the write bandwidth of the memory. Furthermore, in the display memory application where the memory banks or modules (collectively called blocks) can be mapped into different horizontal display lines, a properly set-up broadcast-write can burst-write multiple horizontal lines simultaneously without the need to go through consecutive "page breaks" as in prior art memories.
In the present embodiment, the chip I/O drives the IMB with full CMOS swing, and the MBI drives the IMB with a reduced CMOS swing. The MBI drives the IMB with normal CMOS inverters, but the signal swing is clamped by the receiving circuitry in the chip I/O. FIG. 8 shows an embodiment of this clamping circuit (a feedback circuit, which is a variation of the clamping circuit shown in FIG. 2(a)). A negative feedback circuit (which includes the receiver inverter I2, two inverter-like N/P transistor pairs that stack on one another: transistor P3/N3 and P2/N2 and P1/N1) is used to limit the signal swing and two control transistors P4, N4 in series with transistors P1 and N1.
These transistors together with the disable and disable # signals are for disabling the feedback circuit. The swing limitation circuit is placed in the chip I/O area to limit the signal swing when one of the modules is driving the bus in the direction to the chip I/O. This is advantageous because the use of small swing signals in this direction allows the bus line drivers in the MBI to be significantly smaller compared to the case when full-swing signal is used, and resulting in significant reduction in bus capacitance contributed by the bus drivers since the memory module is replicated 18 times. This reduction in capacitance also translates to reduction in noise generation. In the other direction when the chip I/O is driving, full CMOS swing is used, the receivers in the MBI are normal CMOS inverters (without feedback circuitry); no extra silicon area is dedicated to the feedback circuit in the memory modules.
Since large signal swing requires large driver but small, simple receiver and small swing requires small driver but relatively complicated receiver (receiver with feedback circuit) by using large swing in the chip I/O to memory module direction (one driver and multiple receivers) and small swing in the module to I/O direction (multiple driver and one receiver). Area and power are minimized.
Column Address Generation
Burst mode implementation in the DRAM requires a mechanism that will allow sequentially addressed data to be accessed, given only the starting address of the sequence. Using this starting address, the memory will generates subsequent addresses which are decoded to select the appropriate column lines. An address sequencer is needed in the memory chip to sequence properly the column select lines during a burst mode.
FIG. 9 illustrates a prior art method to sequence the column select lines. The starting address is loaded on address bus 911 into a counter 901 controlled by the memory clock and connected to a decoder 902 by bus 912. The output of this decoder 902 on bus 913 is buffered by buffers 903 to drive the column select lines 0, . . . , 2N. At every rising clock edge, the counter 901 increments and its output is decoded to generate the next active column select line. The column select lines are asserted in consecutive order for the duration of one clock cycle. One drawback to this approach is that the delay time from the clock to valid column select is the sum of the clock to out time of the counter 901, the propagation delays of the decoder 902 and the buffer 903. This delay directly affects the burst frequency and therefore access bandwidth. Because the decoder typically is implemented with combinational logic, the propagation delay through the decoder is not uniform for each output transition. As a result, simultaneous assertion of one or more column selects may occur for the duration of the decoder delay mismatches. This can cause read or write failures during high speed operations.
One embodiment of the present invention uses another approach, shown in FIG. 10(a), which sequences the column select lines. Instead of loading the starting address into a counter, the address on the address bus is fed to a decoder 1001. The decoder 1001 decodes which column select is to be asserted. The outputs from the decoder 1001, which are all logic low except for one logic one, are loaded into a conventional barrel shifter 1002. (The barrel shifter includes registers connected in a ring configuration with the number of stages equal to the number of outputs from the decoder 1001, as shown in FIG. 10(b), which shifts the logic one to the next register at rising edges of the clock.) If the consecutive buffered outputs of the barrel shifter are connected in correspondence to consecutive column select lines 0, . . . , 2N, then the column select lines will be asserted in consecutive order for the duration of one clock cycle. The functional behavior is the same as that of the prior art scheme. One advantage of the present scheme is that the delay time from the clock to valid column select is just the clock-to-out time of the latch plus the buffer delay. The decoder delay is eliminated. Secondly, as long as the decoded outputs are stable before the rising edge of the clock, the barrel shifter output will update in unison, timed to the rising edge of the clock. Since there are no combinational logic gates in the path of clock to column selects, there will be a uniform delay for clock to valid column selects. Simultaneous column select assertion is easily suppressed; therefore, this scheme minimizes read or write failures caused by address transitions.
The full address width (such as eight bits for row, five bits for column in the embodiment described) is usually divided into several subsets, each of which is represented as in FIGS. 9 and 10a, 10b to reduce the number of column select lines (2N). In both the prior art and the present invention, the counter or the barrel shifter can be made to advance in incrementing or decrementing order according to certain direction-control logic (not shown).
Inter-block Burst Mode
With the multiple block architecture, the memory can further implement a novel inter-block burst read and write operation. Instead of bursting data in and out from the column circuitry of a particular block, multiple blocks can participate in the burst read or write operations. Basically, each participating block (through command setup to enter the inter-block mode) latches in or sends out a data word at each clock cycle consecutively as shown in FIG. 11. Memory modules 1101 to 1118 each include a mode-bit which can be set by a special command. Modules 1101 to 1104 with their mode bit set are selected for inter-block burst access operation and accessed in a repeated sequence corresponding to the present base address sequence.
In one embodiment, the bank with its ID matching all 8 bits in the ID field of the command sends first, for instance 1102, followed by 1103, 1104 and 1101 in the corresponding order. After all four banks have finished sending the first word, the CAS address in the address incrementer is increased by 1 and the sending cycle starts over again. This continues until the burst read command is terminated. Note that the burst data is not interrupted during the whole interleave access. This mode is particularly useful when the column sense amplifier caches of participating blocks are mapped into consecutive or different horizontal display lines, and thus the inter-block access represents "vertical" access to the display memory. As a result, both the horizontal access (normal column burst) and the vertical access of the display memory can be extremely fast (at burst rate).
Incorporating synchronous burst access capability into a DRAM operating at high frequencies increases the data access bandwidth; however, this increase in performance comes at the expense of possible increased latency for the first data accessed, increased circuit complexity and die area.
In prior art memory designs, because of the incorporation of pipelines, implementing synchronous burst transfers normally results in increased latency for the first data accessed. In particular, the access latency is limited by the number of pipelines incorporated and is therefore quantized to clock cycles as opposed to traditional asynchronous DRAMs where access latency is measured in time. FIG. 12(a) shows an example of prior art resynchronization for the read data path. Here the input memory clock (Clk) is used to clock the pipeline stage and re-synchronize data out from the memory array. FIG. 12(b) is an associated timing diagram showing the "latency penalty" caused by synchronous burst transfer.
As shown in FIG. 12(a) and 12(b), delay S is the delay from the point the address is clocked into the address sequencer 1202 to the point the data OUT from the buffer 1205 is available at an input IN of a synchronization register 1211. Delay B is the delay from the point the gate G2 is opened by a leading edge of the clock to the point the data is available at the pad 1210. The access latency under synchronous burst transfer as shown in FIG. 12(a) is a clock cycle T. Therefore the latency penalty is the difference between a block cycle T and delay S plus delay B, as shown in FIG. 12(b).
In accordance with the present invention, an innovative method reduces the access latency penalty associated with synchronous burst operation. The DRAM thereby has the capability, at lower operating frequencies, to transmit valid data from the initialization edge.
Instead of using pipeline stages synchronized by the input memory clock (Clk) as describe above, only an address latch is used in the memory bus interface (MBI). In other words, there is no synchronization pipeline through the entire read or write access paths. For command/address or data write operation, where data are sent from the chip I/O to the memory modules, the latch in the module MBI is controlled by a buffered version of the signal Clk. Delay of the buffered clock is matched to the delay of the input data path by using similar layout and capacitive loading. The buffered clock controls the address sequencing and writing to the memory. This is one embodiment of fully-matched source synchronous data and clock path. For read access, the buffered clock is routed back to the chip I/O again fully matched to the read data path to become the Tclk (the source synchronous clock of the read path).
In the chip I/O area, incorporated is a FIFO memory which has an input controlled by the Tclk and an output controlled by the Clk. The FIFO is used for the re-synchronization of the data from Tclk to Clk. FIG. 13 shows the schematic of such a 4-deep FIFO 1300. The FIFO 1300 is specially designed for data input at both edges of the Tclk. Tsel and Tsel# are outputs of a toggle flip-flop 1305 which is controlled by the Tclk. Similarly, Osel and Osel# are outputs of another toggle flip-flop 1306 controlled by Mclk, which is a gated version of Clk. The reset signal is used to reset the flip-flops at the beginning of the transfer cycle. The enable lines are used to enable the state update of the flip-flops. For every Tclk edge, a piece of data is written into a latch inside the FIFO, and the write operation cycles from latches A to D.
FIG. 14(a) shows memory clock Mclk, source synchronous clock signal Tclk, signal Tsel, data signal at the input of FIFO, signals at the outputs of latches A to D, signal Osel and the signals at pad 1304. Edge 0 is assumed to be a initialization edge. The delay between clock signals Mclk and Tclk is delay S+B which in the shown example is less than three half-clock periods. Delay S and B have same meaning as shown in FIG. 12(a).
As seen from FIGS. 13 and 14(a), at any time only one pair of two serial transistors turn on, other pairs turn off. Specifically, before edge 0' of the signal Tclk, only transistors G4 and G5 turn on because both signals Tclk and Tsel# are high. At edge 0' transistors G4 and G45 turn off and transistors G7 and G8 turn on because signals Tclk and Tsel both transition from low to high. Data De latched into latch B is held therein until transistor pair G4 and G5 turn on again at edge 3'. Transistor G7 turns off at edge 1' because its gate signal becomes low and data D1 latched into latch C at time t1 is held therein until both transistors G7 and G8 conduct again at edge 4'.
A similar situation happens to transistor pair G10 and G11 which conduct at edge 1' and turns off-at edge 2', and to transistor pair G1 and G2. The output signals of latches A to D are shown in FIG. 14(a) designating Out A to Out D. Each latch A to D holds a data for about one and half clock cycle in a definitive order. Latch B holds data D0, D4, D8, . . . latch C, D1, D5, D9 . . . ; latch D, D2, D6, D10 . . . ; and latch A, D3, D7, D11 . . . .
Similarly, data is output from FIFO 1300 at each edge of Mclk. A programmable control logic (not shown) is used to control Mclk so that Mclk is turned on at a certain number Nck of half-clock periods after the memory command/address is received. This number Nck is a measure of the total memory access delay (Tmad), the delay of Tclk from the input memory clock Mclk. In cases when Tmad is substantially smaller than half a clock period, Tclk and Mclk are very much in synchronization, Nck is set to zero and the memory operates like an asynchronous memory with no synchronization pipeline. Data flows through the FIFO's without waiting for the input or output gates (G1-G14) to open.
In FIG. 14(a) Nck is equal to 4 and the first data is transmitted out on the memory clock's rising edge. Examples of sampling and update sequence shown in FIG. 14(b) are similar to those shown in FIG. 14(a) except the first data is transmitted on the falling edge and the number Nck in this case is equal to 3.
Although the embodiment shown in FIG. 13 implements data transfer at both clock edges, a similar circuit can be used for re-synchronizing data with single clock edge transfers. One example is to remove the two latches controlled by Tclk# and Mclk# signals in FIG. 13 to form a two entry FIFO with data transfer at rising edge only. More entries may also be added with appropriate modification and expansion to the Tsel and Osel signals. The N-channel transfer gates shown in FIGS. 12(a) and 13 are for illustrated purpose only; those circuits may be implemented using complimentary transfer gate, P-channel gate or other logically equivalent gates.
Prior art memory redundancy circuit activates the "test mode" to test redundant memory sub-array (which may include redundant word-lines, redundant bit-lines, and redundant blocks) through the use of special "tri-level" voltage level on certain input pins (M. Hamada et al, "Semiconductor Memory Apparatus with a Spare Memory Cell Array", U.S. Pat. No. 5,113,371). With the present multiple block architecture, such an arrangement is not flexible.
Referring to FIG. 15(a), in accordance with the present invention, configuration bit or bits ("test" bit(s)) are included in each memory block 1500, so that the redundant memory sub-array(s) 1505, 1506 inside each block can be individually and independently activated and tested when the test bit(s) in test register 1502 are turned-on that enables the redundant sub-array 1505,1506 to be tested while disabling the main array 1507.
FIG. 15(b) shows a construct of the test register 1502 of FIG. 15(a). It consists of three D type flip-flops, output of which enable the different part of memory array 1505, 1506, 1507 for access. The test register 1502 is programmable through IMB 1520 and the assertion of the write enable signal 1530. This allows for in-system memory intra-block testing and repairing. The fuses 1540, 1541 and 1542 are used to permanently enable or disable the corresponding part of the memory array.
This disclosure is illustrative and not limiting; further modifications and variations will be apparent to those skilled in the art in light of this disclosure and the appended claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US3585378 *||Jun 30, 1969||Jun 15, 1971||Ibm||Error detection scheme for memories|
|US3651473 *||Mar 27, 1970||Mar 21, 1972||Burroughs Corp||Expandable interlock exchange for multiprocessing systems|
|US3761879 *||May 8, 1972||Sep 25, 1973||Philips Corp||Bus transport system for selection information and data|
|US3803562 *||Nov 21, 1972||Apr 9, 1974||Honeywell Inf Systems||Semiconductor mass memory|
|US3849872 *||Oct 24, 1972||Nov 26, 1974||Ibm||Contacting integrated circuit chip terminal through the wafer kerf|
|US3983537 *||Apr 1, 1974||Sep 28, 1976||Hawker Siddeley Dynamics Limited||Reliability of random access memory systems|
|US4007452 *||Jul 28, 1975||Feb 8, 1977||Intel Corporation||Wafer scale integration system|
|US4038648 *||Jun 3, 1974||Jul 26, 1977||Chesley Gilman D||Self-configurable circuit structure for achieving wafer scale integration|
|US4063225 *||Mar 8, 1976||Dec 13, 1977||Rca Corporation||Memory cell and array|
|US4071887 *||Oct 30, 1975||Jan 31, 1978||Motorola, Inc.||Synchronous serial data adaptor|
|US4092733 *||May 7, 1976||May 30, 1978||Mcdonnell Douglas Corporation||Electrically alterable interconnection|
|US4188670 *||Jan 11, 1978||Feb 12, 1980||Mcdonnell Douglas Corporation||Associative interconnection circuit|
|US4227045 *||Jun 28, 1978||Oct 7, 1980||Honeywell Inc.||Data processing protocol system|
|US4319356 *||Dec 19, 1979||Mar 9, 1982||Ncr Corporation||Self-correcting memory system|
|US4329685 *||Jun 9, 1980||May 11, 1982||Burroughs Corporation||Controlled selective disconnect system for wafer scale integrated circuits|
|US4379327 *||Jul 21, 1980||Apr 5, 1983||Motorola, Inc.||Universal interface circuit for synchronous and asynchronous buses|
|US4407014 *||Oct 6, 1980||Sep 27, 1983||Honeywell Information Systems Inc.||Communications subsystem having a direct connect clock|
|US4458297 *||Nov 29, 1982||Jul 3, 1984||Mosaic Systems, Inc.||Universal interconnection substrate|
|US4467400 *||May 13, 1982||Aug 21, 1984||Burroughs Corporation||Wafer scale integrated circuit|
|US4494196 *||Sep 7, 1983||Jan 15, 1985||Wang Laboratories, Inc.||Controller for peripheral data storage units|
|US4605928 *||Oct 24, 1983||Aug 12, 1986||International Business Machines Corporation||Fault-tolerant array of cross-point switching matrices|
|US4615017 *||Sep 19, 1983||Sep 30, 1986||International Business Machines Corporation||Memory controller with synchronous or asynchronous interface|
|US4627058 *||Jan 25, 1985||Dec 2, 1986||Pioneer Electronic Corporation||Code error correction method|
|US4630355 *||Mar 8, 1985||Dec 23, 1986||Energy Conversion Devices, Inc.||Electric circuits having repairable circuit lines and method of making the same|
|US4639861 *||Jan 20, 1984||Jan 27, 1987||Cselt Centro Studi E Laboratori Telecomunicazioni S.P.A.||Interface controlling bidirectional data transfer between a synchronous and an asynchronous bus|
|US4646298 *||May 1, 1984||Feb 24, 1987||Texas Instruments Incorporated||Self testing data processing system with system test master arbitration|
|US4653050 *||Dec 3, 1984||Mar 24, 1987||Trw Inc.||Fault-tolerant memory system|
|US4663758 *||Aug 23, 1985||May 5, 1987||Cselt-Centro Studi E Laboratori Telecomunicazioni Spa||Wideband integrated services local communication system|
|US4667328 *||Apr 29, 1985||May 19, 1987||Mieczyslaw Mirowski||Clocking circuit with back-up clock source|
|US4703436 *||Feb 1, 1984||Oct 27, 1987||Inova Microelectronics Corporation||Wafer level integration technique|
|US4707808 *||Apr 26, 1985||Nov 17, 1987||Rockwell International Corporation||Small size, high speed GaAs data latch|
|US4719621 *||Jul 15, 1985||Jan 12, 1988||Raytheon Company||Packet fastbus|
|US4847615 *||Oct 6, 1986||Jul 11, 1989||Anamartic Limited||Control system for chained circuit modules|
|US4866508 *||Sep 26, 1986||Sep 12, 1989||General Electric Company||Integrated circuit packaging configuration for rapid customized design and unique test capability|
|US4872137 *||Nov 21, 1985||Oct 3, 1989||Jennings Iii Earle W||Reprogrammable control circuit|
|US4881232 *||Feb 5, 1988||Nov 14, 1989||Sony Corporation||Method and apparatus for error correction|
|US4906987 *||Dec 15, 1986||Mar 6, 1990||Ohio Associated Enterprises, Inc.||Printed circuit board system and method|
|US4926382 *||Nov 23, 1988||May 15, 1990||Kabushiki Kaisha Toshiba||Divided bit line type dynamic random access memory with charging/discharging current suppressor|
|US4943914 *||Oct 20, 1987||Jul 24, 1990||Hitachi, Ltd.||Storage control system in which real address portion of TLB is on same chip as BAA|
|US4943966 *||Apr 8, 1988||Jul 24, 1990||Wang Laboratories, Inc.||Memory diagnostic apparatus and method|
|US4955020 *||Jun 29, 1989||Sep 4, 1990||Infotron Systems Corporation||Bus architecture for digital communications|
|US4970724 *||Dec 22, 1988||Nov 13, 1990||Hughes Aircraft Company||Redundancy and testing techniques for IC wafers|
|US4974048 *||Mar 10, 1989||Nov 27, 1990||The Boeing Company||Integrated circuit having reroutable conductive paths|
|US4980765 *||Jan 12, 1990||Dec 25, 1990||Hitachi, Ltd.||Frame buffer memory for display|
|US4984192 *||Dec 2, 1988||Jan 8, 1991||Ultrasystems Defense Inc.||Programmable state machines connectable in a reconfiguration switching network for performing real-time data processing|
|US4985895 *||Nov 14, 1988||Jan 15, 1991||Wegener Communications, Inc.||Remote controlled receiving system apparatus and method|
|US5020020 *||Apr 7, 1989||May 28, 1991||Digital Equipment Corporation||Computer interconnect system with transmit-abort function|
|US5043820 *||Mar 27, 1989||Aug 27, 1991||Hughes Aircraft Company||Focal plane array readout employing one capacitive feedback transimpedance amplifier for each column|
|US5045725 *||Mar 30, 1989||Sep 3, 1991||Kabushiki Kaisha Toshiba||Integrated standard cell including clock lines|
|US5055897 *||Jul 27, 1988||Oct 8, 1991||Intel Corporation||Semiconductor cell for neural network and the like|
|US5077596 *||Jul 20, 1990||Dec 31, 1991||Mitsubishi Denki Kabushiki Kaisha||Semiconductor device|
|US5077737 *||Aug 18, 1989||Dec 31, 1991||Micron Technology, Inc.||Method and apparatus for storing digital data in off-specification dynamic random access memory devices|
|US5077738 *||May 22, 1991||Dec 31, 1991||Intel Corporation||Test mode enable scheme for memory|
|US5103424 *||Mar 26, 1990||Apr 7, 1992||Trw Inc.||Memory column interface with fault tolerance|
|US5111271 *||Jun 25, 1990||May 5, 1992||Kabushiki Kaisha Toshiba||Semiconductor device using standard cell system|
|US5111434 *||May 31, 1989||May 5, 1992||Ltd. SamSung Electronics Co.||Semiconductor memory device|
|US5128737 *||Mar 2, 1990||Jul 7, 1992||Silicon Dynamics, Inc.||Semiconductor integrated circuit fabrication yield improvements|
|US5133064 *||Apr 22, 1988||Jul 21, 1992||Hitachi, Ltd.||Data processing system generating clock signal from an input clock, phase locked to the input clock and used for clocking logic devices|
|US5159273 *||Sep 28, 1990||Oct 27, 1992||Hughes Aircraft Company||Tri-state bus driver to support reconfigurable fault tolerant logic|
|US5187779 *||Aug 11, 1989||Feb 16, 1993||Micral, Inc.||Memory controller with synchronous processor bus and asynchronous i/o bus interfaces|
|US5204836 *||Oct 30, 1990||Apr 20, 1993||Sun Microsystems, Inc.||Method and apparatus for implementing redundancy in parallel memory structures|
|US5214657 *||Jun 30, 1992||May 25, 1993||Micron Technology, Inc.||Method for fabricating wafer-scale integration wafers and method for utilizing defective wafer-scale integration wafers|
|US5218686 *||Nov 3, 1989||Jun 8, 1993||Compaq Computer Corporation||Combined synchronous and asynchronous memory controller|
|US5243703 *||Mar 5, 1992||Sep 7, 1993||Rambus, Inc.||Apparatus for synchronously generating clock signals in a data processing system|
|US5252507 *||Mar 30, 1990||Oct 12, 1993||Tactical Fabs, Inc.||Very high density wafer scale device architecture|
|US5254883 *||Apr 22, 1992||Oct 19, 1993||Rambus, Inc.||Electrical current source circuitry for a bus|
|US5261077 *||Jun 29, 1990||Nov 9, 1993||Digital Equipment Corporation||Configurable data path arrangement for resolving data type incompatibility|
|US5268639 *||Jun 5, 1992||Dec 7, 1993||Rambus, Inc.||Testing timing parameters of high speed integrated circuit devices|
|US5329174 *||Oct 23, 1992||Jul 12, 1994||Xilinx, Inc.||Circuit for forcing known voltage on unconnected pads of an integrated circuit|
|EP0178949A2 *||Oct 18, 1985||Apr 23, 1986||Fujitsu Limited||Bipolar-transistor type semiconductor memory device having a redundancy configuration|
|EP0389203A2 *||Mar 16, 1990||Sep 26, 1990||Fujitsu Limited||Semiconductor memory device having information indicative of presence of defective memory cells|
|JPS59212962A *||Title not available|
|WO1991016680A1 *||Apr 16, 1991||Oct 31, 1991||Rambus Inc||Integrated circuit i/o using a high preformance bus interface|
|WO1993018459A1 *||Mar 3, 1993||Sep 16, 1993||Rambus Inc||Prefetching into a cache to minimize main memory access time and cache size in a computer system|
|WO1993018462A1 *||Mar 3, 1993||Sep 16, 1993||Rambus Inc||High speed bus system|
|WO1993018463A1 *||Mar 3, 1993||Sep 16, 1993||Rambus Inc||Method and circuitry for minimizing clock-data skew in a bus system|
|1||"32K x 9 Bit BurstRAM™ Synchronous Static RAM with burst Counter and Self-timed Write," Motorola Memory Data, pp. 7-100 to 7-109.|
|2||"A Design and Yield Evaluation Technique for Wafer-Scale Memory", Apr. 1992, Yamashita et al., Fujitsu.|
|3||"A Monolithic Hough Transform Processor Based on Restructurable VLSI", Rhodes et al., Jan. 1988.|
|4||"Designing With the IDT49C460 and IDT 39C60 Error Detection and Correction Unite", Stodieck, Jan. 1989.|
|5||"Large Area Defect-Tolerant Tree Architectures", Shi et al., 1991.|
|6||"Reconfiguration of Binary Trees: The Flow-Driven Approach", Antola et al., 1991.|
|7||"Task-Flow Architecture for WSI Parallel Processing", Apr. 1992, Robert W. Horst, Tandem Computers, Inc.|
|8||"Wafer-Scale Integration--A Fault-Tolerant Procedure", Aubusson et al., Jun. 1978.|
|9||"Yield Optimization in Large RAM's with Hierarchical Redundancy", Ganapathy et al., Sep. 1991.|
|10||*||32K x 9 Bit BurstRAM Synchronous Static RAM with burst Counter and Self timed Write, Motorola Memory Data, pp. 7 100 to 7 109.|
|11||*||A Design and Yield Evaluation Technique for Wafer Scale Memory , Apr. 1992, Yamashita et al., Fujitsu.|
|12||*||A Monolithic Hough Transform Processor Based on Restructurable VLSI , Rhodes et al., Jan. 1988.|
|13||Cavil et al., "Wafer-Scale Integration," Microelectronics Manufacturing Technology, May, 1991, pp. 55-59.|
|14||*||Cavil et al., Wafer Scale Integration, Microelectronics Manufacturing Technology, May, 1991, pp. 55 59.|
|15||*||Designing With the IDT49C460 and IDT 39C60 Error Detection and Correction Unite , Stodieck, Jan. 1989.|
|16||Fujii, et al., "A 50-μA Standby 1Mx1/256Kx4 CMOS DRAM with High-Speed Sense Amplifier," IEEE Jnl of Solid State Circuits, vol. SC-21, No. 5, Oct. 1986, pp. 643-647.|
|17||*||Fujii, et al., A 50 A Standby 1Mx1/256Kx4 CMOS DRAM with High Speed Sense Amplifier, IEEE Jnl of Solid State Circuits, vol. SC 21, No. 5, Oct. 1986, pp. 643 647.|
|18||Herbert Stopper, "Wafer-Scale Integration," Hybrids and High-Level Integration, pp. 354-364.|
|19||*||Herbert Stopper, Wafer Scale Integration, Hybrids and High Level Integration, pp. 354 364.|
|20||*||Large Area Defect Tolerant Tree Architectures , Shi et al., 1991.|
|21||MacDonald et al, "Dynamic RAMs 200mb Wafer Memory," IEEE ISSCC, Feb. 17, 1989, pp. 240-241 and 350.|
|22||*||MacDonald et al, Dynamic RAMs 200mb Wafer Memory, IEEE ISSCC, Feb. 17, 1989, pp. 240 241 and 350.|
|23||*||Peter van Zant, Microchip, A Practical Guide to Semiconductor Processing, 1st Ed., Semiconductor Services, San Jose, CA 1986, p. 8.|
|24||*||Reconfiguration of Binary Trees: The Flow Driven Approach , Antola et al., 1991.|
|25||Ron Iscoff, "Characterizing Quickturn ASICs: It's Done with Mirrors" semiconductor International, Aug. 1, 1990, pp. 68-73.|
|26||*||Ron Iscoff, Characterizing Quickturn ASICs: It s Done with Mirrors semiconductor International, Aug. 1, 1990, pp. 68 73.|
|27||Takai et al., "250 Mbyte/s Synchronous DRAM Using a 3-Stage-Pipelined Architecture," IEEE Jnl of Solid State circuirts, vol. 29, No. 4, Apr. 1994, pp. 426-431.|
|28||*||Takai et al., 250 Mbyte/s Synchronous DRAM Using a 3 Stage Pipelined Architecture, IEEE Jnl of Solid State circuirts, vol. 29, No. 4, Apr. 1994, pp. 426 431.|
|29||*||Task Flow Architecture for WSI Parallel Processing , Apr. 1992, Robert W. Horst, Tandem Computers, Inc.|
|30||*||Wafer Scale Integration A Fault Tolerant Procedure , Aubusson et al., Jun. 1978.|
|31||*||Yield Optimization in Large RAM s with Hierarchical Redundancy , Ganapathy et al., Sep. 1991.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US5706474 *||May 31, 1995||Jan 6, 1998||Hitachi, Ltd.||Synchronous memory system with asynchronous internal memory operation|
|US5715476 *||Dec 29, 1995||Feb 3, 1998||Intel Corporation||Method and apparatus for controlling linear and toggle mode burst access sequences using toggle mode increment logic|
|US5812418 *||Oct 31, 1996||Sep 22, 1998||International Business Machines Corporation||Cache sub-array method and apparatus for use in microprocessor integrated circuits|
|US5838177 *||Jan 6, 1997||Nov 17, 1998||Micron Technology, Inc.||Adjustable output driver circuit having parallel pull-up and pull-down elements|
|US5844913 *||Apr 4, 1997||Dec 1, 1998||Hewlett-Packard Company||Current mode interface circuitry for an IC test device|
|US5870347 *||Mar 11, 1997||Feb 9, 1999||Micron Technology, Inc.||Multi-bank memory input/output line selection|
|US5872736 *||Oct 28, 1996||Feb 16, 1999||Micron Technology, Inc.||High speed input buffer|
|US5910920 *||Dec 18, 1997||Jun 8, 1999||Micron Technology, Inc.||High speed input buffer|
|US5917758 *||Nov 4, 1996||Jun 29, 1999||Micron Technology, Inc.||Adjustable output driver circuit|
|US5920518 *||Feb 11, 1997||Jul 6, 1999||Micron Technology, Inc.||Synchronous clock generator including delay-locked loop|
|US5923594 *||Feb 17, 1998||Jul 13, 1999||Micron Technology, Inc.||Method and apparatus for coupling data from a memory device using a single ended read data path|
|US5926047 *||Aug 29, 1997||Jul 20, 1999||Micron Technology, Inc.||Synchronous clock generator including a delay-locked loop signal loss detector|
|US5940608 *||Feb 11, 1997||Aug 17, 1999||Micron Technology, Inc.||Method and apparatus for generating an internal clock signal that is synchronized to an external clock signal|
|US5946244 *||Mar 5, 1997||Aug 31, 1999||Micron Technology, Inc.||Delay-locked loop with binary-coupled capacitor|
|US5949254 *||Nov 26, 1996||Sep 7, 1999||Micron Technology, Inc.||Adjustable output driver circuit|
|US5953284 *||Jul 9, 1997||Sep 14, 1999||Micron Technology, Inc.||Method and apparatus for adaptively adjusting the timing of a clock signal used to latch digital signals, and memory device using same|
|US5956502 *||Mar 5, 1997||Sep 21, 1999||Micron Technology, Inc.||Method and circuit for producing high-speed counts|
|US5959481 *||Feb 18, 1997||Sep 28, 1999||Rambus Inc.||Bus driver circuit including a slew rate indicator circuit having a one shot circuit|
|US5982679 *||May 29, 1998||Nov 9, 1999||Sgs-Thomson Microelectronics S.A.||Memory circuit with dynamic redundancy|
|US5994942 *||Nov 12, 1997||Nov 30, 1999||Kabushiki Kaisha Toshiba||Buffer circuit with wide dynamic range|
|US6011732 *||Aug 20, 1997||Jan 4, 2000||Micron Technology, Inc.||Synchronous clock generator including a compound delay-locked loop|
|US6014759 *||Jun 13, 1997||Jan 11, 2000||Micron Technology, Inc.||Method and apparatus for transferring test data from a memory array|
|US6016282 *||May 28, 1998||Jan 18, 2000||Micron Technology, Inc.||Clock vernier adjustment|
|US6021512 *||Nov 27, 1996||Feb 1, 2000||International Business Machines Corporation||Data processing system having memory sub-array redundancy and method therefor|
|US6026050 *||Feb 10, 1999||Feb 15, 2000||Micron Technology, Inc.||Method and apparatus for adaptively adjusting the timing of a clock signal used to latch digital signals, and memory device using same|
|US6026456 *||Dec 15, 1995||Feb 15, 2000||Intel Corporation||System utilizing distributed on-chip termination|
|US6029250 *||Sep 9, 1998||Feb 22, 2000||Micron Technology, Inc.||Method and apparatus for adaptively adjusting the timing offset between a clock signal and digital signals transmitted coincident with that clock signal, and memory device and system using same|
|US6035381 *||Dec 12, 1996||Mar 7, 2000||Hitachi, Ltd.||Memory device including main memory storage and distinct key storage accessed using only a row address|
|US6044429 *||Jul 10, 1997||Mar 28, 2000||Micron Technology, Inc.||Method and apparatus for collision-free data transfers in a memory device with selectable data or address paths|
|US6069504 *||Sep 3, 1998||May 30, 2000||Micron Technnology, Inc.||Adjustable output driver circuit having parallel pull-up and pull-down elements|
|US6084434 *||Jun 9, 1999||Jul 4, 2000||Micron Technology, Inc.||Adjustable output driver circuit|
|US6091646 *||Jun 18, 1999||Jul 18, 2000||Micron Technology, Inc.||Method and apparatus for coupling data from a memory device using a single ended read data path|
|US6101197 *||Sep 18, 1997||Aug 8, 2000||Micron Technology, Inc.||Method and apparatus for adjusting the timing of signals over fine and coarse ranges|
|US6115318 *||Dec 3, 1996||Sep 5, 2000||Micron Technology, Inc.||Clock vernier adjustment|
|US6122217 *||Feb 4, 1999||Sep 19, 2000||Micron Technology, Inc.||Multi-bank memory input/output line selection|
|US6138205 *||Aug 28, 1998||Oct 24, 2000||Nec Corporation||Burst mode type semiconductor memory device|
|US6173432||Jun 20, 1997||Jan 9, 2001||Micron Technology, Inc.||Method and apparatus for generating a sequence of clock signals|
|US6195395 *||Mar 18, 1998||Feb 27, 2001||Intel Corporation||Multi-agent pseudo-differential signaling scheme|
|US6201424||May 20, 1999||Mar 13, 2001||Micron Technology, Inc.||Synchronous clock generator including a delay-locked loop signal loss detector|
|US6226205 *||Feb 22, 1999||May 1, 2001||Stmicroelectronics, Inc.||Reference voltage generator for an integrated circuit such as a dynamic random access memory (DRAM)|
|US6256255||Jul 17, 2000||Jul 3, 2001||Micron Technology, Inc.||Multi-bank memory input/output line selection|
|US6256259||May 12, 2000||Jul 3, 2001||Micron Technology, Inc.||Delay-locked loop with binary-coupled capacitor|
|US6262921||May 12, 2000||Jul 17, 2001||Micron Technology, Inc.||Delay-locked loop with binary-coupled capacitor|
|US6269451||Feb 27, 1998||Jul 31, 2001||Micron Technology, Inc.||Method and apparatus for adjusting data timing by delaying clock signal|
|US6272608||Apr 28, 1999||Aug 7, 2001||Micron Technology, Inc.||Method and apparatus for synchronous data transfers in a memory device with lookahead logic for detecting latency intervals|
|US6279090||Sep 3, 1998||Aug 21, 2001||Micron Technology, Inc.||Method and apparatus for resynchronizing a plurality of clock signals used in latching respective digital signals applied to a packetized memory device|
|US6279116||Sep 3, 1999||Aug 21, 2001||Samsung Electronics Co., Ltd.||Synchronous dynamic random access memory devices that utilize clock masking signals to control internal clock signal generation|
|US6313659 *||Sep 22, 2000||Nov 6, 2001||Sun Microsystems, Inc.||Controlled impedance CMOS receiver for integrated circuit communication between circuits|
|US6326810||May 21, 1999||Dec 4, 2001||Micron Technology, Inc.||Adjustable output driver circuit|
|US6327196||Mar 24, 2000||Dec 4, 2001||Micron Technology, Inc.||Synchronous memory device having an adjustable data clocking circuit|
|US6338127||Aug 28, 1998||Jan 8, 2002||Micron Technology, Inc.||Method and apparatus for resynchronizing a plurality of clock signals used to latch respective digital signals, and memory device using same|
|US6340904||May 24, 1999||Jan 22, 2002||Micron Technology, Inc.||Method and apparatus for generating an internal clock signal that is synchronized to an external clock signal|
|US6343036 *||Sep 18, 1998||Jan 29, 2002||Samsung Electronics Co., Ltd.||Multi-bank dynamic random access memory devices having all bank precharge capability|
|US6349399||Sep 3, 1998||Feb 19, 2002||Micron Technology, Inc.||Method and apparatus for generating expect data from a captured bit pattern, and memory device using same|
|US6374360||Dec 11, 1998||Apr 16, 2002||Micron Technology, Inc.||Method and apparatus for bit-to-bit timing correction of a high speed memory bus|
|US6378079||Mar 16, 2000||Apr 23, 2002||Micron Technology, Inc.||Computer system having memory device with adjustable data clocking|
|US6400641||Jul 15, 1999||Jun 4, 2002||Micron Technology, Inc.||Delay-locked loop with binary-coupled capacitor|
|US6405280||Jun 5, 1998||Jun 11, 2002||Micron Technology, Inc.||Packet-oriented synchronous DRAM interface supporting a plurality of orderings for data block transfers within a burst sequence|
|US6415340||Sep 15, 2000||Jul 2, 2002||Micron Technology, Inc.||Method and apparatus for synchronous data transfers in a memory device with selectable data or address paths|
|US6430696||Nov 30, 1998||Aug 6, 2002||Micron Technology, Inc.||Method and apparatus for high speed data capture utilizing bit-to-bit timing correction, and memory device using same|
|US6437600||Aug 7, 2001||Aug 20, 2002||Micron Technology, Inc.||Adjustable output driver circuit|
|US6438063||Nov 16, 2000||Aug 20, 2002||Samsung Electronics Co., Ltd.||Integrated circuit memory devices having selectable column addressing and methods of operating same|
|US6477675||Aug 7, 2001||Nov 5, 2002||Micron Technology, Inc.||Method and apparatus for generating expect data from a captured bit pattern, and memory device using same|
|US6483757||Jun 29, 2001||Nov 19, 2002||Micron Technology, Inc.||Delay-locked loop with binary-coupled capacitor|
|US6490207||Nov 2, 2001||Dec 3, 2002||Micron Technology, Inc.||Delay-locked loop with binary-coupled capacitor|
|US6490224||Jul 16, 2001||Dec 3, 2002||Micron Technology, Inc.||Delay-locked loop with binary-coupled capacitor|
|US6499111||Jul 16, 2001||Dec 24, 2002||Micron Technology, Inc.||Apparatus for adjusting delay of a clock signal relative to a data signal|
|US6519719||Jan 6, 2000||Feb 11, 2003||Micron Technology, Inc.||Method and apparatus for transferring test data from a memory array|
|US6556483||Oct 9, 2001||Apr 29, 2003||Micron Technology, Inc.||Method and apparatus for synchronous data transfers in a memory device with selectable data or address paths|
|US6560668||Oct 9, 2001||May 6, 2003||Micron Technology, Inc.||Method and apparatus for reading write-modified read data in memory device providing synchronous data transfers|
|US6567338||Jul 25, 2000||May 20, 2003||Integrated Device Technology, Inc.||Fully synchronous pipelined RAM|
|US6591354||May 26, 1999||Jul 8, 2003||Integrated Device Technology, Inc.||Separate byte control on fully synchronous pipelined SRAM|
|US6611885||Oct 9, 2001||Aug 26, 2003||Micron Technology, Inc.||Method and apparatus for synchronous data transfers in a memory device with selectable data or address paths|
|US6614698||Oct 9, 2001||Sep 2, 2003||Micron Technology, Inc.|
|US6643787 *||Oct 19, 1999||Nov 4, 2003||Rambus Inc.||Bus system optimization|
|US6643789||Jan 23, 2002||Nov 4, 2003||Micron Technology, Inc.||Computer system having memory device with adjustable data clocking using pass gates|
|US6647523||Oct 9, 2002||Nov 11, 2003||Micron Technology, Inc.||Method for generating expect data from a captured bit pattern, and memory device using same|
|US6662304||Jan 14, 2002||Dec 9, 2003||Micron Technology, Inc.||Method and apparatus for bit-to-bit timing correction of a high speed memory bus|
|US6690191 *||Dec 21, 2001||Feb 10, 2004||Sun Microsystems, Inc.||Bi-directional output buffer|
|US6785188||Jan 28, 2002||Aug 31, 2004||Integrated Device Technology, Inc.||Fully synchronous pipelined RAM|
|US6789175||Oct 9, 2001||Sep 7, 2004||Micron Technology, Inc.|
|US6801989||Jun 28, 2001||Oct 5, 2004||Micron Technology, Inc.||Method and system for adjusting the timing offset between a clock signal and respective digital signals transmitted along with that clock signal, and memory device and computer system using same|
|US6934202||Jan 16, 2003||Aug 23, 2005||Sgs-Thomson Microelectronics S.A.||Memory circuit with dynamic redundancy|
|US6950956||Nov 3, 2003||Sep 27, 2005||Rambus Inc.||Integrated circuit with timing adjustment mechanism and method|
|US6965539||Dec 19, 2003||Nov 15, 2005||Hynix Semiconductor Inc.||Write path scheme in synchronous DRAM|
|US6982922||Mar 11, 2003||Jan 3, 2006||Rambus Inc.||Single-clock, strobeless signaling system|
|US6990042||Feb 10, 2005||Jan 24, 2006||Rambus Inc.||Single-clock, strobeless signaling system|
|US7042914||Oct 13, 2003||May 9, 2006||Rambus Inc.||Calibrated data communication system and method|
|US7050342 *||Sep 10, 2003||May 23, 2006||Micron Technology, Inc.||Testmode to increase acceleration in burn-in|
|US7069406||Jul 2, 1999||Jun 27, 2006||Integrated Device Technology, Inc.||Double data rate synchronous SRAM with 100% bus utilization|
|US7159067 *||Nov 7, 2003||Jan 2, 2007||Hitachi, Ltd.||Information processing apparatus using index and TAG addresses for cache|
|US7280419 *||Nov 3, 2006||Oct 9, 2007||Nanya Technology Corp.||Latency counter having frequency detector and latency counting method thereof|
|US7320082||Dec 18, 2003||Jan 15, 2008||Rambus Inc.||Power control system for synchronous memory device|
|US7397725||Dec 30, 2005||Jul 8, 2008||Rambus Inc.||Single-clock, strobeless signaling system|
|US7535933||Jan 5, 2006||May 19, 2009||Rambus Inc.||Calibrated data communication system and method|
|US7574616||Sep 17, 2004||Aug 11, 2009||Rambus Inc.||Memory device having a power down exit register|
|US7581121||Jun 14, 2005||Aug 25, 2009||Rambus Inc.||System for a memory device having a power down mode and method|
|US7626880||Apr 15, 2005||Dec 1, 2009||Rambus Inc.||Memory device having a read pipeline and a delay locked loop|
|US7657813||Apr 18, 2008||Feb 2, 2010||Micron Technology, Inc.||Method and apparatus for generating expect data from a captured bit pattern, and memory device using same|
|US7663966||Jul 1, 2008||Feb 16, 2010||Rambus, Inc.||Single-clock, strobeless signaling system|
|US7889593||Aug 31, 2007||Feb 15, 2011||Round Rock Research, Llc||Method and apparatus for generating a sequence of clock signals|
|US7954031||Dec 29, 2009||May 31, 2011||Round Rock Research, Llc|
|US7986584||Oct 29, 2009||Jul 26, 2011||Rambus Inc.||Memory device having multiple power modes|
|US7996590 *||Aug 12, 2009||Aug 9, 2011||Samsung Electronics Co., Ltd.||Semiconductor memory module and semiconductor memory system having termination resistor units|
|US8102730||Feb 16, 2010||Jan 24, 2012||Rambus, Inc.||Single-clock, strobeless signaling system|
|US8107580||Sep 18, 2009||Jan 31, 2012||Round Rock Research, Llc||Method and apparatus for generating a phase dependent control signal|
|US8127152||Nov 19, 2004||Feb 28, 2012||Rambus Inc.||Method of operation of a memory device and system including initialization at a first frequency and operation at a second frequency and a power down exit mode|
|US8170067||May 1, 2012||Rambus Inc.||Memory system with calibrated data communication|
|US8181092||Dec 15, 2006||May 15, 2012||Round Rock Research, Llc||Dynamic synchronization of data capture on an optical or other high speed communications link|
|US8248884||Dec 21, 2010||Aug 21, 2012||Rambus Inc.||Method of controlling a memory device having multiple power modes|
|US8305839||Jan 17, 2012||Nov 6, 2012||Rambus Inc.||Memory device having multiple power modes|
|US8433023||Jan 20, 2012||Apr 30, 2013||Round Rock Research, Llc||Method and apparatus for generating a phase dependent control signal|
|US8565008||Feb 4, 2011||Oct 22, 2013||Round Rock Research, Llc||Method and apparatus for generating a sequence of clock signals|
|US8630317||Apr 13, 2012||Jan 14, 2014||Rambus Inc.||Memory system with calibrated data communication|
|US8756395||Feb 27, 2012||Jun 17, 2014||Rambus Inc.||Controlling DRAM at time DRAM ready to receive command when exiting power down|
|US8892974||May 14, 2012||Nov 18, 2014||Round Rock Research, Llc||Dynamic synchronization of data capture on an optical or other high speed communications link|
|US8948212||Jan 13, 2014||Feb 3, 2015||Rambus Inc.||Memory controller with circuitry to set memory device-specific reference voltages|
|US9164933||Feb 3, 2015||Oct 20, 2015||Rambus Inc.||Memory system with calibrated data communication|
|US20010024135 *||Jan 9, 2001||Sep 27, 2001||Harrison Ronnie M.||Method and apparatus for generating a sequence of clock signals|
|US20030167374 *||Jul 2, 1999||Sep 4, 2003||Stanley A. Hronik||Double data rate synchronous sram with 100% bus utilization|
|US20030231537 *||Mar 11, 2003||Dec 18, 2003||Stark Donald C.||Single-clock, strobeless signaling system|
|US20040017692 *||Jan 16, 2003||Jan 29, 2004||Sgs-Thomson Microelectronics S.A.||Memory circuit with dynamic redundancy|
|US20040047226 *||Sep 10, 2003||Mar 11, 2004||Gans Dean D.||Testmode to increase acceleration in burn-in|
|US20040093458 *||Nov 7, 2003||May 13, 2004||Hitachi, Ltd.||Information processing apparatus using index and TAG addresses for cache|
|US20040098634 *||Nov 3, 2003||May 20, 2004||Zerbe Jared Levan||Integrated circuit with timing adjustment mechanism and method|
|US20040103226 *||Oct 14, 2003||May 27, 2004||Brian Johnson||Method and system for adjusting the timing offset between a clock signal and respective digital signals transmitted along with clock signal, and memory device and computer system using same|
|US20040141404 *||Dec 18, 2003||Jul 22, 2004||Rambus Inc.||Power control system for synchronous memory device|
|US20040158785 *||Aug 25, 2003||Aug 12, 2004||Manning Troy A.|
|US20050047264 *||Dec 19, 2003||Mar 3, 2005||Hynix Semiconductor Inc.||Write path scheme in synchronous DRAM|
|US20050060487 *||Sep 17, 2004||Mar 17, 2005||Barth Richard M.||Memory device having a power down exit register|
|US20050091464 *||Oct 27, 2003||Apr 28, 2005||Ralph James||System and method for using a learning sequence to establish communications on a high-speed nonsynchronous interface in the absence of clock forwarding|
|US20050120161 *||Nov 19, 2004||Jun 2, 2005||Barth Richard M.||Methods of operation of a memory device and system|
|US20050141335 *||Feb 10, 2005||Jun 30, 2005||Rambus Inc.||Single-clock, strobeless signaling system|
|US20050180255 *||Apr 15, 2005||Aug 18, 2005||Tsern Ely K.||Memory device having a read pipeline and a delay locked loop|
|US20050216654 *||May 25, 2005||Sep 29, 2005||Barth Richard M||System and module including a memory device having a power down mode|
|US20050235130 *||Jun 14, 2005||Oct 20, 2005||Barth Richard M||System for a memory device having a power down mode and method|
|US20050249028 *||Jul 15, 2005||Nov 10, 2005||Harrison Ronnie M||Method and apparatus for generating a sequence of clock signals|
|US20050286505 *||Aug 12, 2005||Dec 29, 2005||Harrison Ronnie M||Method and apparatus for generating a phase dependent control signal|
|US20060103453 *||Oct 26, 2005||May 18, 2006||Stmicroelectronics S.R.L.||Voltage down converter|
|US20060104151 *||Dec 30, 2005||May 18, 2006||Rambus Inc.||Single-clock, strobeless signaling system|
|US20060120409 *||Jan 5, 2006||Jun 8, 2006||Zerbe Jared L||Calibrated data communication system and method|
|US20060206742 *||May 11, 2006||Sep 14, 2006||Ralph James||System and method for using a learning sequence to establish communications on a high- speed nonsynchronous interface in the absence of clock forwarding|
|US20080126059 *||Aug 31, 2007||May 29, 2008||Micron Technology, Inc.||Method and apparatus for generating a sequence of clock signals|
|US20080267000 *||Jul 1, 2008||Oct 30, 2008||Rambus Inc.||Single-clock, strobeless signaling system|
|US20080279323 *||Jul 23, 2008||Nov 13, 2008||Harrison Ronnie M||Method and apparatus for generating a phase dependent control signal|
|US20080301533 *||Dec 15, 2006||Dec 4, 2008||Micron Technology, Inc.||Dynamic synchronization of data capture on an optical or other high speed communications link|
|US20090303802 *||Aug 12, 2009||Dec 10, 2009||Samsung Electronics Co., Ltd.||Semiconductor memory module and semiconductor memory system having termination resistor units|
|US20090327789 *||Dec 31, 2009||Zerbe Jared Levan||Memory System with Calibrated Data Communication|
|US20100046314 *||Feb 25, 2010||Tsern Ely K||Memory Device Having a Read Pipeline and a Delay Locked Loop|
|US20150187435 *||Dec 30, 2013||Jul 2, 2015||Unisys Corporation||Systems and methods for memory management in a dynamic translation computer system|
|EP0768671A2 *||Oct 11, 1996||Apr 16, 1997||Cirrus Logic, Inc.||Circuits, systems and methods for reducing power loss during transfer of data across a conductive line|
|U.S. Classification||327/323, 714/E11.009, 365/189.06, 327/309, 257/E21.526, 327/319, 711/E12.086|
|International Classification||G06F5/16, H01L27/02, G11C29/00, G06F11/20, G06F11/00, G06F11/10, H01L21/66, G06F13/40, G11C29/48, G06F12/06, H04L25/02|
|Cooperative Classification||Y02B60/1235, Y02B60/1225, Y02B60/1228, G06F13/4077, H01L22/22, G11C29/832, H01L27/0203, G06F11/006, G06F11/10, G11C29/76, G11C29/48, G06F12/0661, H04L25/026, G06F5/16, G11C29/808, G11C29/88, G11C29/006, G06F11/1666|
|European Classification||G11C29/88, G11C29/808, G11C29/76, H01L22/22, G11C29/832, G06F5/16, G06F13/40E2B2, G11C29/00W, G11C29/48, G06F12/06K2D, H04L25/02G|
|Oct 29, 1996||CC||Certificate of correction|
|Aug 23, 1999||FPAY||Fee payment|
Year of fee payment: 4
|Aug 1, 2003||FPAY||Fee payment|
Year of fee payment: 8
|Sep 4, 2007||FPAY||Fee payment|
Year of fee payment: 12
|Apr 3, 2008||AS||Assignment|
Owner name: MOSYS, INC., CALIFORNIA
Free format text: CHANGE OF NAME;ASSIGNOR:MONOLITHIC SYSTEM TECHNOLOGY, INC.;REEL/FRAME:020741/0975
Effective date: 20060525
Owner name: MOSYS, INC.,CALIFORNIA
Free format text: CHANGE OF NAME;ASSIGNOR:MONOLITHIC SYSTEM TECHNOLOGY, INC.;REEL/FRAME:020741/0975
Effective date: 20060525